Multiple Choice

Multi-tenant SaaS with 500 customer-specific fine-tunes. Merge or swap?

Why are small FT datasets especially vulnerable to membership inference?

When does targeting all-linear with LoRA pay off vs. just W_q+W_v?

LoRA+: why does B want a higher learning rate than A?

LoRA init: why B = 0 and A ~ Gaussian (not both Gaussian)?

Stacking two unrelated LoRA adapters: most likely outcome

If you change LoRA's rank from 16 to 64 with α fixed, what actually happens?

Why does fine-tuning on company docs often raise hallucination instead of lowering it?

DPO vs RLHF: pick the most accurate operational difference

DoRA: which component does LoRA actually adapt?

Distillation with a closed-API teacher: which method is feasible?

Why does pure long-context FT (without RoPE scaling) fail to extrapolate?

Which signal is the most reliable indicator of catastrophic forgetting?

When is gradient accumulation NOT equivalent to a real larger batch?

A chatbot team reports TTFT P99 jumped from 300ms to 1.4s overnight. Which root cause is most likely?

Which mitigation has the largest impact on agent-loop latency when tool calls are independent?

Which workload most benefits from RadixAttention specifically vs simpler per-request prefix caching?

On the roofline, which intervention moves an LLM decode workload most directly toward higher peak throughput?

Which scenario most justifies the extra cost of QAT over PTQ for inference quantization?

Which prompt-construction pattern is most likely to actually hit the provider's prompt cache?

Which statement best explains why prefill and decode hit different bottlenecks on the same GPU?

What is the primary mechanism by which PagedAttention increases serving throughput?

An engineer says 'FP8 KV cache just halves memory, no other benefit.' Which response is most correct?

Which statement most accurately captures why GQA, not MQA, became the production default?

Where does FlashAttention deliver the biggest serving speedup?